Semantic Maps for Knowledge Graphs: A Semantic-Based Summarization Approach

Pablo Camarillo-Ramirez,Francisco Cervantes-Alvarez,Luis Fernando Gutiérrez-Preciado

DOI: https://doi.org/10.1109/access.2024.3351170

IF: 3.9

2024-01-01

IEEE Access

Abstract:Knowledge Graphs (KGs) have emerged as a powerful tool for representing semantic structured information and enabling the development of intelligent systems. This paper focuses on the generation of semantic maps as summarization method for KGs. We propose a strategy that utilizes centroid-based clustering algorithms, namely Affinity Propagation and Partitioning Around Medoids (PAM), to capture the semantic distance between nodes in the KG and generate meaningful clusters. Our experiments demonstrate divergent results between the two clustering algorithms, with Affinity Propagation showing qualitative coherence and meaningfulness, while PAM performs well in terms of internal validation metrics. We leverage the computed centroids to infer a main term of the semantic map, which contributes to the visually appealing and informative representation of the KG. The combination of semantic distance capture, clustering algorithms, and centroid-based inference facilitates a comprehensive understanding of the KG. Our findings highlight the importance of considering both qualitative and quantitative evaluation measures in assessing clustering results. The effectiveness of semantic maps is showcased in visualizing KGs and advancing the field of knowledge graph visualization. The integration of centroid-based clustering algorithms, qualitative evaluation, and inference methods offers improved clarity and interpretability for complex KG analysis.

computer science, information systems,telecommunications,engineering, electrical & electronic